In backpointers fsck, we do a seqential scan of one btree, and check
references to another: extents <-> backpointers
Checking references generates random lookups, so we want to pin that
btree in memory (or only a range, if it doesn't fit in ram).
Previously, this was done with a simple check in the shrinker - "if
btree node is in range being pinned, don't free it" - but this generated
OOMs, as our shrinker wasn't well behaved if there was less memory
available than expected.
Instead, we now have two different shrinkers and lru lists; the second
shrinker being for pinned nodes, with seeks set much higher than normal
- so they can still be freed if necessary, but we'll prefer not to.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
32 bits won't overflow any time soon, but size_t is the correct type for
counting objects in memory.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
"bcachefs; Fix deadlock in bch2_btree_update_start()" was a significant
performance regression (nearly 50%) on multithreaded random writes with
fio.
The reason is that the journal watermark checks multiple things,
including the state of the btree write buffer, and on multithreaded
update heavy workloads we're bottleneked on write buffer flushing - we
don't want kicknig off btree updates to depend on the state of the write
buffer.
This isn't strictly correct; the interior btree update path does do
write buffer updates, but it's a tiny fraction of total accounting
updates and we're more concerned with space in the journal itself.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
If a journal write errored, the list of devices it was written to could
be empty - we're not supposed to mark an empty replicas list.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This prevents going emergency read only when the user has specified
replicas_required > replicas.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Previosuly, the transaction commit path would have to add keys to the
btree write buffer as a separate operation, requiring additional global
synchronization.
This patch introduces a new journal entry type, which indicates that the
keys need to be copied into the btree write buffer prior to being
written out. We switch the journal entry type back to
JSET_ENTRY_btree_keys prior to write, so this is not an on disk format
change.
Flushing the btree write buffer may require pulling keys out of journal
entries yet to be written, and quiescing outstanding journal
reservations; we previously added journal->buf_lock for synchronization
with the journal write path.
We also can't put strict bounds on the number of keys in the journal
destined for the write buffer, which means we might overflow the size of
the preallocated buffer and have to reallocate - this introduces a
potentially fatal memory allocation failure. This is something we'll
have to watch for, if it becomes an issue in practice we can do
additional mitigation.
The transaction commit path no longer has to explicitly check if the
write buffer is full and wait on flushing; this is another performance
optimization. Instead, when the btree write buffer is close to full we
change the journal watermark, so that only reservations for journal
reclaim are allowed.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Ensure that journal bufs that haven't been written can't be reclaimed
from the journal pin fifo, and can thus have new pins taken.
Prep work for changing the btree write buffer to pull keys from the
journal directly.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This introduces a new helper for connecting time_stats to state changes,
i.e. when taking journal reservations is blocked for some reason.
We use this to track separately the different reasons the journal might
be blocked - i.e. space in the journal full, or the journal pin fifo
full.
Also do some cleanup and improvements on the time stats code.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Previously, bch2_journal_pin_set() would silently ignore a request to
pin a journal sequence number that was no longer dirty, because it was
used internally by bch2_journal_pin_copy() which could race with the src
pin being flushed.
Split these apart so that we can properly assert that @seq is a
currently dirty journal sequence number - this is almost always a bug.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Since outstanding journal buffers hold a journal pin, when flushing all
pins we need to close the current journal entry if necessary so its pin
can be released.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This deletes the complicated and somewhat expensive journal
pre-reservation machinery in favor of just using journal watermarks:
when the journal is more than half full, we run journal reclaim more
aggressively, and when the journal is more than 3/4s full we only allow
journal reclaim to get new journal reservations.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We have a couple journal pin put helpers to handle cases where the
journal lock is already held or not. Refactor the helpers to lock
and reclaim from the highest level and open code the reclaim from
the one caller of the internal variant. The latter call will be
moved into the journal buf release helper in a later patch.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Split out a new file for bch_sb_field_members - we'll likely want to
move more code here in the future.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
We need to always call bch2_replicas_gc_end() after we've called
bch2_replicas_gc_start(), else we leave state around that needs to be
cleaned up.
Partial fix for: https://github.com/koverstreet/bcachefs/issues/560
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
- endianness fixes
- mark some things static
- fix a few __percpu annotations
- fix silent enum conversions
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
A simple device evacuate, remove, add test loop with concurrent
shutdowns occasionally reproduces a problem where the filesystem
fails to mount. The mount failure occurs because the filesystem was
uncleanly shut down, yet no member device is marked for journal data
in the superblock. An fsck detects the problem, restores the mark
and allows the mount to proceed without further consistency issues.
The reason for the lack of journal data marks is the gc mechanism
invoked via bch2_journal_flush_device_pins() runs while the journal
happens to be empty. This results in garbage collection of all journal
replicas entries. Once the updated replicas table is written to the
superblock, the filesystem is put in a transiently unrecoverable state
until further journal data is written, because journal recovery expects
to find at least one marked journal device whenever the filesystem is
not otherwise marked clean (i.e. as on clean unmount).
To fix this problem, update the journal replicas gc algorithm to always
mark currently active journal replicas entries by writing to the
journal. This ensures that only entries for devices that are no longer
used for journaling are garbage collected, not just those that don't
happen to currently hold journal data. This preserves the journal
recovery invariant above and avoids putting the fs into a transiently
unrecoverable state.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
GFP_NOIO dates from the bcache days, when we operated under the block
layer. Now, GFP_NOFS is more appropriate, so switch all GFP_NOIO uses to
GFP_NOFS.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
The journal stucking check in bch2_journal_space_available() is
particularly aggressive and can lead to premature shutdown in some
rare cases. This is difficult to reproduce, but also comes along
with a fatal error and so is worthwhile to be cautious.
For example, we've seen instances where the journal is under heavy
reservation pressure, the journal allocation path transitions into
the final available journal bucket, the journal write path
immediately consumes that bucket and calls into
bch2_journal_space_available(), which then in turn flags the journal
as stuck because there is no available space and shuts down the
filesystem instead of submitting the journal write (that would have
otherwise succeeded).
To avoid this problem, simplify the journal stuck checking by just
relying on the higher level logic in the journal reservation path.
This produces more useful debug output and is a more reliable
indicator that things have bogged down.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Tweak journal reclaim to ensure the btree node cache isn't more
than half dirty so that memory reclaim can always make progress - the
same as we do for the btree key cache.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
It's now legal for the pin fifo to be empty, which means this code needs
to be updated in order to not hit an assert.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
checkpatch.pl gives lots of warnings that we don't want - suggested
ignore list:
ASSIGN_IN_IF
UNSPECIFIED_INT - bcachefs coding style prefers single token type names
NEW_TYPEDEFS - typedefs are occasionally good
FUNCTION_ARGUMENTS - we prefer to look at functions in .c files
(hopefully with docbook documentation), not .h
file prototypes
MULTISTATEMENT_MACRO_USE_DO_WHILE
- we have _many_ x-macros and other macros where
we can't do this
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Also, do some reorganizing/renaming, convert atomic counters in bch_fs
to persistent counters, and add a few missing counters.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Delete some obsolete tracepoints, organize alloc tracepoints better,
make a few tracepoints more consistent.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Since journal reclaim -> btree key cache flushing may require the
allocation of new btree nodes, it has an implicit dependency on copygc
in order to make forward progress - so we should avoid blocking copygc
unless the journal is really close to full.
This introduces watermarks to replace our single MAY_GET_UNRESERVED bit
in the journal, and adds a watermark for copygc and plumbs it through.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
When bch2_journal_pin_set() is updating an existing pin, we shouldn't
call bch2_journal_reclaim_fast() after dropping the old pin and before
dropping the new pin - that could reclaim the entry we're trying to pin.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
bch2_journal_space_available -> bch2_journal_halt() self deadlocks on
journal lock; work around this by dropping/retaking journal lock before
we call bch2_fatal_error().
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
It makes the code more readable if we work off of sequence numbers,
instead of direct indexes into the array of journal buffers.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This patch changes journal_entry_open() to initialize the new journal
entry, not __journal_entry_close().
This also means that journal_cur_seq() refers to the sequence number of
the last journal entry when we don't have an open journal entry, not the
next one.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
journal_flush_done() was overwriting did_work, thus occasionally
returning false when it did do work and occasional assertions in the
shutdown sequence because we didn't completely flush the key cache.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
This patch changes printbufs dynamically allocate and reallocate a
buffer as needed. Stack usage has become a bit of a problem, and a major
cause of that has been static size string buffers on the stack.
The most involved part of this refactoring is that printbufs must now be
exited with printbuf_exit().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This patch was originally to work around the journal geting stuck in
nochanges mode - but that was just a hack, we needed to fix the actual
bug. It should be fixed now, so revert it.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
When the nochanges option is selected, we're supposed to never issue
writes. Unfortunately, it seems discards were missed when implemnting
this, leading to some painful filesystem corruption.
Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
